A BOOSTING APPROACH FOR INTRUSION DETECTION

在线阅读 下载PDF 导出详情
摘要 Intrusiondetectioncanbeessentiallyregardedasaclassificationproblem,namely,dis-tinguishingnormalprofilesfromintrusivebehaviors.Thispaperintroducesboostingclassificationalgorithmintotheareaofintrusiondetectiontolearnattacksignatures.Decisiontreealgorithmisusedassimplebaselearnerofboostingalgorithm.Furthermore,thispaperemploysthePrincipleCom-ponentAnalysis(PCA)approach,aneffectivedatareductionapproach,toextractthekeyattributesetfromtheoriginalhigh-dimensionalnetworktrafficdata.KDDCUP99datasetisusedintheseex-perimentstodemonstratethatboostingalgorithmcangreatlyimprovetheclassificationaccuracyofweaklearnersbycombininganumberofsimple“weaklearners”.Inourexperiments,theerrorrateoftrainingphaseofboostingalgorithmisreducedfrom30.2%to8%after10iterations.Besides,thispaperalsocomparesboostingalgorithmwithSupportVectorMachine(SVM)algorithmandshowsthattheclassificationaccuracyofboostingalgorithmislittlebetterthanSVMalgorithm’s.However,thegeneralizationabilityofSVMalgorithmisbetterthanboostingalgorithm.
机构地区 不详
出版日期 2007年03月13日(中国期刊网平台首次上网日期,不代表论文的发表时间)
  • 相关文献